Topic
  • 1 reply
  • Latest Post - ‏2013-04-02T19:01:18Z by lisaw
SystemAdmin
SystemAdmin
1485 Posts

Pinned topic Client loader optimization

‏2013-03-15T14:53:51Z |
Hi
I in my POC project for extreme scale have used clien tloader. The underneath JPA provider is hibernate .
I have planning to cache big tables with 4 -6 millions or more in extreme scale .For small tables everything was running fin e.
Now in order to get an optimum results , I wanted to change some parameter settings.

I see the hibernate queeries getting fired underneath . I caluclated them and found it fires 1 querry to load 1000 rows. I now want to increase the no of rows fetched per query .Since i have 4 million rows , it will be optimal for me to load more roes per querry => less number of querries=> less round trips to DB.
I tried changing th ejdbc fetch size in persistence .xml but no gains
can anyone suggest me how to tune these parameter in client loader.

Thanks
Pritish
Updated on 2013-04-02T19:01:18Z at 2013-04-02T19:01:18Z by lisaw
  • lisaw
    lisaw
    101 Posts

    Re: Client loader optimization

    ‏2013-04-02T19:01:18Z  
    I believe setMaxResults(1000) is the default if not specified, that's why you see that behavior.

    If you are using the API to preload for example, you can set maxResults or setPreloadBatchSize() ObjectGridHibernateCacheProvider API.

    For write behind, I think there is a property you can set in the persistence.xml
    Configuring the Hibernate cache plug-in

    wxs.write_behind_max_batch_size=10000

    Then when you query, you'll get back more than 1000 results.

    Hope that helps!

    • Lisa

    Websphere eXtreme Scale Development